U.S. patent application number 14/783291 was filed with the patent office on 2016-03-24 for inspection apparatus.
The applicant listed for this patent is BRIDGESTONE CORPORATION. Invention is credited to Yoshitaka FUJISAWA.
Application Number | 20160086320 14/783291 |
Document ID | / |
Family ID | 51791978 |
Filed Date | 2016-03-24 |
United States Patent
Application |
20160086320 |
Kind Code |
A1 |
FUJISAWA; Yoshitaka |
March 24, 2016 |
INSPECTION APPARATUS
Abstract
An inspection apparatus capable of preventing a conforming
article from being judged as nonconforming when inspecting a molded
object for acceptability by performing image processing on an image
captured of the inspection object. The apparatus includes a means
for positioning a model pattern in a position with a highest degree
of agreement by matching the model pattern against the image
captured of the inspection object, a model pattern dividing means
for dividing the model pattern into a plurality of elements in such
a manner as to have mutually overlapping regions, and a shape
recognition means for recognizing a shape corresponding to the
model pattern by positioning each of the elements divided by the
model pattern dividing means in a position with a highest degree of
agreement by performing pattern matching within a predetermined
range with reference to the position where the model pattern has
been positioned on the image.
Inventors: |
FUJISAWA; Yoshitaka;
(Kodaira-shi, JP) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
BRIDGESTONE CORPORATION |
Tokyo |
|
JP |
|
|
Family ID: |
51791978 |
Appl. No.: |
14/783291 |
Filed: |
April 25, 2014 |
PCT Filed: |
April 25, 2014 |
PCT NO: |
PCT/JP2014/061664 |
371 Date: |
November 20, 2015 |
Current U.S.
Class: |
382/141 |
Current CPC
Class: |
G06T 7/001 20130101;
G06K 9/4604 20130101; G06K 9/6211 20130101; G06T 2207/20112
20130101; G06T 7/74 20170101; G01N 2201/12 20130101; G01N 21/95607
20130101; G06T 7/11 20170101; G06T 2207/30108 20130101; G06K 9/6201
20130101; G01B 11/24 20130101; G01N 21/8851 20130101; G06K
2009/4666 20130101; G06T 7/13 20170101 |
International
Class: |
G06T 7/00 20060101
G06T007/00; G01B 11/24 20060101 G01B011/24; G06K 9/46 20060101
G06K009/46; G01N 21/88 20060101 G01N021/88 |
Foreign Application Data
Date |
Code |
Application Number |
Apr 25, 2013 |
JP |
2013-092438 |
Claims
1. An inspection apparatus for judging acceptability of an
inspection object by matching a model pattern representing
conforming object data against an image captured of an appearance
of the inspection object, the apparatus comprising: a model pattern
positioning means for positioning the model pattern in a position
with a highest degree of agreement by matching the model pattern
against the image captured; a model pattern dividing means for
dividing the model pattern into a plurality of elements in such a
manner as to have mutually overlapping regions; and a shape
recognition means for recognizing a shape corresponding to the
model pattern by positioning each of the elements divided by the
model pattern dividing means in a position with a highest degree of
agreement by performing pattern matching within a predetermined
range with reference to the position where the model pattern has
been positioned on the image.
2. The inspection apparatus according to claim 1, wherein the model
pattern dividing means performs division of the model pattern a
plurality of times and changes the number of divisions at each time
of division so that the positions of division boundaries of
elements divided at each time of division are dislocated from the
other.
3. The inspection apparatus according to claim 1, wherein the shape
recognition means is provided with an element deforming means to
carry out necessary deformation of the elements, including scaling
deformation, rotational deformation, and shear deformation, and
repeats the deformation of the elements by performing pattern
matching against the image until the shape reaches a highest degree
of agreement.
4. The inspection apparatus according to claim 2, wherein the shape
recognition means is provided with an element deforming means to
carry out necessary deformation of the elements, including scaling
deformation, rotational deformation, and shear deformation, and
repeats the deformation of the elements by performing pattern
matching against the image until the shape reaches a highest degree
of agreement.
Description
TECHNICAL FIELD
[0001] The present invention relates to an inspection apparatus
capable of preventing a conforming article from being judged
nonconforming when inspecting a molded product for acceptability by
performing image processing on an image captured of the appearance
of the inspection object.
BACKGROUND ART
[0002] An inspection apparatus known in the art determines the
acceptability of an inspection object by storing conforming article
data, as a model pattern, of the inspection object, such as molded
characters on a product, and calculating the degree of agreement
between the image pattern obtained of the inspection object and the
model pattern (see Patent Document 1).
[0003] However, in such an inspection, some products, which are in
fact conforming units, may show low degrees of agreement between
the image pattern of the inspection object and the model pattern.
Such situations may sometimes lead to judgment of conforming
articles as nonconforming. For example, let us assume an inspection
for acceptability of the molded state of a character an inspection
for acceptability of the molded state of a character string, which
is an inspection object with protrusions and recesses molded on the
side of the tire. Tires before such inspection are placed in a
certain location temporarily, and, as a result, the sides of tires
can get deformed by the sagging of the rubber or the overlap of
belts. ?? With a character string molded on the side of a tire
deformed as described above, there will be drops in the degree of
agreement of the model pattern with the image pattern of the
character string. Consequently, it may be possible that the
character string molded correctly on the side of a tire is judged
as nonconforming.
PRIOR ART DOCUMENT
Patent Document
[0004] Patent Document 1: Japanese Unexamined Patent Application
Publication No. 2006-275952
SUMMARY OF THE INVENTION
Problems to Be Solved By the Invention
[0005] The purpose of this invention is therefore to provide an
inspection apparatus capable of preventing a conforming article
from being judged nonconforming when inspecting a molded object for
acceptability by performing image processing on an image captured
of the inspection object.
Means for Solving the Problem
[0006] Thus, an inspection apparatus to solve the above-described
problem includes a model pattern positioning means for positioning
a model pattern in a position with a highest degree of agreement by
matching the model pattern against the image captured of the
appearance of an inspection object, a model pattern dividing means
for dividing the model pattern into a plurality of elements in such
a manner as to have mutually overlapping regions, and a shape
recognition means for recognizing a shape corresponding to the
model pattern by positioning each of the elements divided by the
model pattern dividing means in a position with a highest degree of
agreement by performing pattern matching within a predetermined
range with reference to the position where the model pattern has
been positioned on the image. As a result, it is possible to
perform pattern matching of elements, which are the divisions of
the model pattern, accurately with the image captured even when the
image of the inspection object, which is in fact a conforming
article, has been captured with deformity. Accordingly, it is
possible to prevent a conforming inspection object from being
judged as nonconforming because the model pattern is not well
matched to the image.
BRIEF DESCRIPTION OF THE DRAWINGS
[0007] FIG. 1 is a schematic configuration diagram showing a block
constitution of an inspection apparatus.
[0008] FIG. 2 is diagrams showing relationship of an image pattern
and a model pattern and the division of a model pattern.
[0009] FIG. 3 is diagrams showing a matching of elements to an
image pattern.
[0010] FIG. 4 is a diagram showing another mode of division of the
model pattern.
BEST MODE FOR CARRYING OUT THE INVENTION
First Embodiment
[0011] FIG. 1 is a schematic configuration diagram of an inspection
apparatus 1 for a tire 2 implementing an embodiment of the present
invention. As shown in FIG. 1, the inspection apparatus 1 according
to the first embodiment is an apparatus for inspecting for
acceptability each of molded characters, which constitute a
character string 4 with protrusions and recesses molded on the side
3 of the tire 2. The tire 2 in this case is supposed to be one
having a deformed side 3 because it has been placed after
manufacture temporarily in a certain location before the
inspection. The inspection apparatus 1 has an inspection object
imaging unit 5 for capturing an image of the character string 4 on
the side 3 of a "not inflated" tire 2 and a pattern matching unit
6.
[0012] The inspection object imaging unit 5 comprises a unit for
capturing the whole image of the character string 4 by optical
cutting method (light-section method), for instance. The unit
consists of a laser unit 7 for emitting a laser beam at the
character string 4 and a camera 8 for imaging the character string
4 irradiated with the laser beam. Thus an image of the side 3 of
the tire 2 including the character string 4 molded thereon is
captured.
[0013] The pattern matching unit 6 includes an image pattern
generating means 11, a model pattern storage 12 for storing a model
pattern M of the character string 4, and a matching means 13.
[0014] The image pattern generating means 11 derives the shape data
of the character string 4 from the image captured by the inspection
object imaging unit 5 and at the same time generates an image
pattern R from the feature points of the shape data necessary for
recognizing the shape of each character in the character string
4.
[0015] The matching means 13 includes a model pattern positioning
means 15 for positioning the model pattern M in a position with a
highest degree of agreement by matching the model pattern M
indicated by a dotted line against an image pattern R indicated by
a solid line on an image as shown in FIG. 2A, a model pattern
dividing means 16 for dividing the model pattern M into a plurality
of elements M1 to M1 in such a manner as to have mutually
overlapping regions D as shown in FIG. 2B, and a shape recognition
means 17 for recognizing a shape corresponding to the model pattern
M from the image by positioning each of the elements M1 to M4
divided by the model pattern dividing means 16 in a position with a
highest degree of agreement by performing pattern matching within a
predetermined range with reference to the position where the model
pattern has been positioned on the image.
[0016] The above-described means 11, 13, and 15 through 17 are
realized by a computer and programs to have the computer execute
processing procedures by the means 11, 13, and 15 through 17 to be
discussed later.
[0017] The model pattern positioning means 15 performs pattern
matching by moving in pixels the model pattern M of an entire
character string 4 on the image captured by the inspection object
imaging unit 5. In doing so, the model pattern positioning means 15
calculates the degrees of agreement of the model pattern M with the
image pattern R on the image on which the model pattern M is placed
and positions the model pattern M in a position with the highest
degree of agreement with the image pattern R.
[0018] This model pattern M is positioned on the image with the
median point of the model pattern M as a reference position, for
instance. That is, the model pattern M is set on the image by
searching for the image pattern R with the highest degree of
agreement with the model pattern M from the image.
[0019] The model pattern dividing means 16 divides the model
pattern M into a plurality of elements M1 to M4 in equal size in
such a manner as to have mutually overlapping regions D as shown in
FIGS. 2A and 2B. The overlapping regions D are so set as to create
an overlap of a predetermined number of pixels for the neighboring
elements M1 to M4. The overlapping regions D provided for the
divided elements M1 to M4 in this way can prevent gaps from
occurring between the neighboring elements M1 to M4 when the
elements M1 to M4 are matched against the image pattern R. That is,
the setting of overlapping regions D can prevent a mistaken
judgment of nonconformity, which can occur when there are gaps
between the neighboring elements M1 to M4.
[0020] The shape recognition means 17 includes an element deforming
means 18 and an element positioning means 19.
[0021] The element deforming means 18 performs predetermined
deformations on the elements M1 to M4. More specifically, as shown
in FIG. 3A, the element deforming means 18 performs scaling
deformation, rotational deformation, shear deformation, or
deformation combining them in predetermined proportions on the
elements M1 to M4.
[0022] The element positioning means 19 performs pattern matching
against the image pattern R by deforming the elements M1 to M4 by
the element deforming means 18. In this pattern matching, all the
above-mentioned deformations on the elements M1 to M4 may be
repeated by the element deforming means 18 until a shape showing a
highest degree of agreement is obtained. The degree of agreement is
calculated as the position of each of the elements M1 to M4 is
changed in pixels within a predetermined range with reference to
the positions of the elements M1 to M4 when the model pattern M has
been positioned against the image pattern R. Following this, the
shape recognition means 17 performs positioning of each of the
elements M1 to M4 against the image pattern R through further
pattern matching of each of the elements M1 to M4 against the image
pattern R. Thus, the shape recognition means 17 determines whether
the image pattern R has a shape corresponding to the model pattern
M. In this manner, pattern matching against the image pattern R is
performed by deforming the elements M1 to M4 such that the elements
M1 to M4 can be matched to the image pattern R with improved
accuracy. In other words, the elements can be matched to the
inspection object with better accuracy.
[0023] A description is given of the operation of the matching
means 13.
[0024] The model pattern positioning means 15 positions the model
pattern M on the image in a position of the image pattern R where
the degree of agreement with the model pattern M is the highest.
?
[0025] Next, the elements M1 to M4 of the model pattern M divided
by the model pattern dividing means 16 are individually subjected
to shape deformation such as scaling deformation, shear
deformation, rotational deformation, or a combination of them by
the element deforming means 18. Through this, a pattern matching is
performed within a predetermined search range with reference to the
position where the model pattern M has been positioned. And the
elements M1 to M4 are arranged on the image pattern R in the shape
of the elements M1 to M4 showing the highest degree of agreement
with the image pattern R and are recognized as the respective parts
of a character corresponding to the model pattern M.
[0026] For example, let us assume that the character "A", indicated
by a dotted line, of the character string 4 of the model pattern M
positioned on the image by the model pattern positioning means 15
is positioned on the "A", indicated by a solid line, of the image
pattern R.
[0027] In this case, the element positioning means 19 calculates
the degree of agreement of the element M1 constituting a part of
the character "A" whenever the "A" of the character string 4 of the
model pattern M is moved in pixels from the median point within the
search range. Further, as the element deforming means 18 continues
to deform the element M1, calculations are performed of the degree
of agreement of the element M1 of the "A" of the model pattern M
with the image pattern R in the image position where the element M1
is overlapped with the image pattern R. And the image pattern R in
the image position and shape showing the highest degree of
agreement with the element M1 of the model pattern M is recognized
as the character element corresponding to the element M1 of the
model pattern M. This process is repeated in the same way for the
elements M2 to M4 to recognize the image pattern R (character)
corresponding to the model pattern M.
[0028] As described above, the degree of agreement can be enhanced
by improving the accuracy of positioning the elements M1 to M4
divided by the model pattern dividing means 16 on the image pattern
R even when the side 3 of the tire 2 is deformed. It is now
possible to match the model pattern M against the image pattern R
with accuracy. Thus a mistaken judgment of the character string 4
on the side 3 as nonconforming due to a failure to match the model
pattern M with the image pattern R can be prevented even though the
character string 4 is, in fact, correctly molded on the side 3 and
is therefore conforming for actual use.
Second Embodiment
[0029] The first embodiment has been so described that the model
pattern M is divided into 2.times.2 elements by the model pattern
dividing means 16 as shown in FIG. 2A. However, the arrangement may
be such that pattern matching against the image pattern R is
performed by dividing the model pattern M into 3.times.3 elements
N1 to N9 as shown in FIG. 4, in addition to the above number of
divisions, in such a manner that the division boundaries do not
coincide with those of 2.times.2 divisions. In other words, the
model pattern dividing means 16 in the second embodiment divides
the model pattern M into the elements M1 to M4 and the elements N1
to N9, which are different in size from each other. And the shape
corresponding to the model pattern M is recognized by the shape
recognition means 17 whenever the model pattern M is divided in a
different size. In this case, the arrangement may be such that the
shape recognized in the 2.times.2 size is superposed on the shape
recognized in the 3.times.3 size and the shape recognition means 17
recognizes the superposed portions as the shape corresponding to
the model pattern M. In this manner, the elements in different
sizes are combined for shape recognition, with the result that the
model pattern M can be matched to the image pattern R with greater
accuracy. Thus a mistaken judgment of the character string 4
correctly molded on the side 3 as nonconforming can be prevented
when it is actually a conforming article.
[0030] That is, the model pattern dividing means 16 performs
division of the model pattern M a plurality of times. And the model
pattern dividing means 16 changes the number of divisions after
each time of division such that the positions of the division
boundaries of the elements M1 to M4 and the elements N1 to N9
divided at each time of division are dislocated from each other. In
this way, the probability of matching some of the elements divided
in different sizes with the image can be improved.
[0031] Thus, when the model pattern M is divided using the elements
M1 to M4 and the elements N1 to N9, which are different in size
from each other, it is of primary importance that the positions of
the division boundaries of the elements divided at each time of
division are dislocated from each other. Dislocation of the
positions of the division boundaries ensures that the division
boundaries of the model pattern M are supplemented by the elements
divided in a different size. When the element positioning means 19
allocates the elements on the image pattern R, the whole of the
image pattern R may be covered by the elements divided in different
sizes. In other words, it is possible to better associate the model
pattern M with the image pattern R. Therefore, a mistaken judgment
of the character string 4 on the side 3 as nonconforming because of
the occurrence of missing portions where the model pattern M cannot
be allocated or the failure to match the model pattern M to the
image pattern R can be prevented with better accuracy when the
character string 4 is actually conforming.
[0032] In the second embodiment, the model pattern M divided in
different sizes is matched against the image pattern R. Thus, by
employing as the image pattern R only the portions where the
elements in different sizes are superposed on each other, such as
the portions where two pixels or more are superposed, the
inspection can be performed with greater accuracy from this image
pattern R by judging the character string 4 not correctly molded as
nonconforming and the character string 4 correctly molded as
conforming.
[0033] It is to be noted that the numbers of divisions of the model
pattern M in the first embodiment and the second embodiment are
only examples and may be altered as appropriate. The description
thus far has dealt with the inspection of a character string 4
molded on the side 3 of a tire 2. However, the inspection object of
this invention may be any pattern or the like formed on a product
surface.
DESCRIPTION OF REFERENCE NUMERALS
[0034] 1 inspection apparatus [0035] 5 inspection object imaging
unit [0036] 6 pattern matching nit [0037] 11 image pattern
generating means [0038] 12 model pattern storage [0039] 13 matching
means [0040] 15 model pattern positioning means [0041] 16 model
pattern dividing means [0042] 17 shape recognition means [0043] 18
element deforming means [0044] 19 element positioning means
* * * * *